Thanks for the link to the old HP document but I'm already very familiar with how to use a spectrum analyser to measure noise figure as I've done this many times both at my place of work and here at home 
Note that I'm not saying your test method isn't scientific 'because' you are using a spectrum analyser. It is because of the way you appear to be using it. Your test method introduces a whole load of causes of measurement uncertainty and is far from scientific.
It's hard enough to measure NF accurately using an analyser even if you do it with expensive test gear because of the cumulative effect of the various causes of measurement uncertainty. But your method takes this to a whole new level of uncertainty because you seem to be relying on making a gain/noise comparison with a noise level at -125dBm against a signal that is at -25dBm and way into the AGC range of the CB radio. So you get the uncertainty of your ability to measure (by eye within a fraction of 1dB) where the true FSD of the cheap CB meter is, you also get the uncertainty of the gain and AGC accuracy of a CB radio over time plus you get the uncertainty of the band noise with the aerial connected, plus you get the fact you are coupling into the collector of an RF amplfier (highish impedance point) with a low impedance spectrum analyser so you would appear to be loading the amplifier stage in an unnatural way. You also get the uncertainty of the spectrum analyser because it will suffer drift in gain over time and also it's accuracy across such a huge range of signal levels will not be that good.
So this all makes it difficult to decide if a change in noise level on your analyser is caused by a change in gain or by a change in noise figure or some subtle loading change at TP13 (or all three). So your result could be hard to interpret correctly.
What model number HP analyser are you using by the way?
Well first up, I'm pretty clued up on RF design and how to make decent test measurements but I'm certainly not an expert on trends in band noise across the radio spectrum
However, anyone can go online and see the test results of various government agencies that measure the various contributors to the noise levels we see on the bands today. Galactic noise is very significant at 27MHz and is typically 18-20dB above thermal noise.
So I don't think I'm confusing anything here? I'm just stating that the band noise on the 11m band is dominated by galactic noise (in the absence of man made noise) and it's hard to avoid receiving this noise with an outdoor CB aerial.
There is an equation that crudely fits the typical line/curve of galactic noise vs frequency when using an omni aerial.
NFm (dB) = 52 − 23 logF
Where F is the frequency in MHz.
So for 27MHz the result would be 19dB excess band noise due to galactic noise. There may be times when this noise gets masked to a degree by the ionosphere but I have no data to suggest how often this happens or the degree of masking. However, in my experience the noise levels on the CB band have always been high with respect to thermal noise level so getting a lower noise figure on test gear is of limited benefit once you connect an aerial.
